Line data Source code
1 : /**
2 : * Copyright (c) 2025 Huawei Technologies Co., Ltd.
3 : * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
4 : * CANN Open Software License Agreement Version 2.0 (the "License").
5 : * Please refer to the License for details. You may not use this file except in compliance with the License.
6 : * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
7 : * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
8 : * See LICENSE in the root of the software repository for the full text of the License.
9 : */
10 : #ifndef USER_REMOTE_MEM_GETTER_H
11 : #define USER_REMOTE_MEM_GETTER_H
12 :
13 : #include <cstdint>
14 : #include <vector>
15 : #include <array>
16 : #include <string>
17 : #include <functional>
18 : #include "hccl_mem_defs.h"
19 : #include "hccl/hccl_types.h"
20 : #include "log.h"
21 :
22 : constexpr uint32_t MAX_BUFFER_NUM = 30000;
23 :
24 : namespace Hccl {
25 :
26 : template <typename T>
27 : struct RemoteMemCtx {
28 : bool& cacheValid;
29 : std::vector<T>& rmtBufferVec;
30 : std::vector<CommMem>& remoteUserMems;
31 : std::vector<std::string>& memInfoCopies;
32 : std::vector<char*>& memInfoPointers;
33 : CommMem** remoteMem;
34 : char*** memInfos;
35 : uint32_t* memNum;
36 :
37 28 : RemoteMemCtx(
38 : bool& cacheValid, std::vector<T>& rmtBufferVec, std::vector<CommMem>& remoteUserMems,
39 : std::vector<std::string>& memInfoCopies, std::vector<char*>& memInfoPointers, CommMem** remoteMem,
40 : char*** memInfos, uint32_t* memNum)
41 28 : : cacheValid(cacheValid),
42 28 : rmtBufferVec(rmtBufferVec),
43 28 : remoteUserMems(remoteUserMems),
44 28 : memInfoCopies(memInfoCopies),
45 28 : memInfoPointers(memInfoPointers),
46 28 : remoteMem(remoteMem),
47 28 : memInfos(memInfos),
48 28 : memNum(memNum) {};
49 : };
50 :
51 0 : inline HcclMemType CommMemTypeToHcclMemType(CommMemType type)
52 : {
53 0 : switch (type) {
54 0 : case COMM_MEM_TYPE_DEVICE:
55 0 : return HCCL_MEM_TYPE_DEVICE;
56 0 : case COMM_MEM_TYPE_HOST:
57 0 : return HCCL_MEM_TYPE_HOST;
58 0 : default:
59 0 : return HCCL_MEM_TYPE_NUM;
60 : }
61 : }
62 :
63 19 : inline CommMemType HcclMemTypeToCommMemType(HcclMemType type)
64 : {
65 19 : switch (type) {
66 10 : case HCCL_MEM_TYPE_DEVICE:
67 10 : return COMM_MEM_TYPE_DEVICE;
68 4 : case HCCL_MEM_TYPE_HOST:
69 4 : return COMM_MEM_TYPE_HOST;
70 5 : default:
71 5 : return COMM_MEM_TYPE_INVALID;
72 : }
73 : }
74 :
75 : template <typename T>
76 28 : HcclResult GetRemoteUserMems(RemoteMemCtx<T>& remoteMemCtx)
77 : {
78 28 : CHK_PRT_RET(!remoteMemCtx.remoteMem, HCCL_ERROR("[GetRemoteUserMems] remoteMem is nullptr"), HCCL_E_PARA);
79 25 : CHK_PRT_RET(!remoteMemCtx.memInfos, HCCL_ERROR("[GetRemoteUserMems] memInfos is nullptr"), HCCL_E_PARA);
80 22 : CHK_PRT_RET(!remoteMemCtx.memNum, HCCL_ERROR("[GetRemoteUserMems] memNum is nullptr"), HCCL_E_PARA);
81 18 : *(remoteMemCtx.remoteMem) = nullptr;
82 18 : *(remoteMemCtx.memInfos) = nullptr;
83 18 : *(remoteMemCtx.memNum) = 0;
84 18 : uint32_t userMemCount = remoteMemCtx.rmtBufferVec.size();
85 18 : if (userMemCount == 0) {
86 8 : HCCL_WARNING("[%s] bufferNum is 0.", __func__);
87 6 : return HCCL_SUCCESS;
88 : }
89 : // 检查是否有缓存
90 12 : if (!remoteMemCtx.cacheValid) {
91 11 : remoteMemCtx.remoteUserMems.clear();
92 11 : remoteMemCtx.memInfoCopies.clear();
93 11 : remoteMemCtx.memInfoCopies.reserve(userMemCount);
94 11 : remoteMemCtx.memInfoPointers.clear();
95 11 : remoteMemCtx.memInfoPointers.reserve(userMemCount);
96 32 : for (uint32_t i = 0; i < userMemCount; ++i) {
97 19 : auto& rmtBuffer = remoteMemCtx.rmtBufferVec[i];
98 19 : if (rmtBuffer == nullptr) {
99 0 : return HCCL_E_PTR;
100 : }
101 19 : CommMem mem{};
102 19 : mem.type = HcclMemTypeToCommMemType(rmtBuffer->GetMemType());
103 19 : mem.addr = reinterpret_cast<void*>(rmtBuffer->GetAddr());
104 19 : mem.size = rmtBuffer->GetSize();
105 19 : remoteMemCtx.remoteUserMems.push_back(mem);
106 19 : std::string memInfoCopy = rmtBuffer->GetMemInfo();
107 19 : remoteMemCtx.memInfoCopies.push_back(std::move(memInfoCopy));
108 19 : remoteMemCtx.memInfoPointers.push_back(const_cast<char*>(remoteMemCtx.memInfoCopies.back().c_str()));
109 23 : HCCL_INFO(
110 : "[%s] Found buffer[addr:%p, size:%llu, memInfo:%s]", __func__, mem.addr, mem.size,
111 : remoteMemCtx.memInfoCopies.back().c_str());
112 : }
113 11 : remoteMemCtx.cacheValid = true;
114 : }
115 12 : *(remoteMemCtx.remoteMem) = remoteMemCtx.remoteUserMems.data();
116 12 : *(remoteMemCtx.memInfos) = remoteMemCtx.memInfoPointers.data();
117 12 : *(remoteMemCtx.memNum) = userMemCount;
118 12 : return HCCL_SUCCESS;
119 : }
120 : } // namespace Hccl
121 :
122 : #endif // USER_REMOTE_MEM_GETTER_H
|